Performance monitoring of large scale parallel computers creates a dilemma: we need to collect detailed information to find performance bottlenecks, yet collecting all this data can introduce serious data collection bottlenecks. At the same time, users are being inundated with volumes of complex graphs and tables that require a performance expert to interpret. We present a new approach called the W 3 Search Model, that addresses both these problems by combining dynamic on-the-fly selection of what performance data to collect with decision support to assist users with the selection and presentation of performance data. We present a case study describing how a prototype implementation of our technique was able to identify the bottlenecks in three real programs. In addition, we were able to reduce the amount of performance data collected by a factor ranging from 13 to 700 compared to traditional sampling and trace based instrumentation techniques. 1. Introduction Performance monitorin...

In distributed systems the need arises to provide event notification based on specified criteria. The criteria change as requests are added or withdrawn. In this paper we describe an approach to the implementation of event notification systems which are based on the theory of finite state automata. 1 Introduction A problem which arises in the context of distributed systems may be characterised as follows-- create an event notification system which is able to detect any one of a number of almost arbitrarily specified events, and notify the requestor. It is assumed that the number of outstanding requests at any instant may be moderately large, say in the tens to hundreds, but varies only slowly over time, as requests are added or withdrawn. The candidate events, on the other hand, occur at an extremely high rate, corresponding perhaps to individual packet transmissions on a network [5]. Any solution to this problem which attempts to sequentially pattern-match candidate events against ea...

this paper we describe the design and implementation of an integrated monitoring and debugging system for a distributed real-time computer system. The monitor provides continuous, transparent monitoring capabilities throughout a real-time system's lifecycle with bounded, minimal, predictable interference by using software support. The monitor is flexible enough to observe both high-level events that are operating system- and application-specific, as well as low-level events such as shared variable references. We present a novel approach to monitoring shared variable references that provides transparent monitoring with low overhead. The monitor is designed to support tasks such as debugging realtime applications, aiding real-time task scheduling, and measuring system performance. Since debugging distributed real-time applications is particularly difficult, we describe how the monitor can be used to debug distributed and parallel applications by deterministic execution replay

. This paper deals with performance evaluation of parallel and distributed systems based on monitoring of concurrent interdependent activities. First a model is introduced for describing the dynamic behavior of computer systems in terms of events. Then, a distributed hardware/hybrid monitor system based on event driven monitoring and its tool environment SIMPLE are presented. We emphasize the tool environment as a prerequisite for successful performance evaluation. The tool environment for evaluating event traces, which integrates the data access interface TDL/POET and a set of evaluation tools for processing the data, makes evaluation independent of the monitor device(s) and the object system. It provides a problem oriented way of accessing event traces. 1. Introduction The characteristic feature of parallel and distributed computer systems is that they share load and common resources among several processing nodes in order to increase performance and reliability of the overall syste...

An approach to sensor configuration, installation and activation for real-time monitoring of broadband networks for managing its performance is presented. An objective-driven measurement strategy for establishing the dynamic and statistical databases of the network is described. Objective driven monitoring allows the activation of sensors for data collection and abstraction based on a set of objectives. The objectives are derived from the quality of service requirements for real-time traffic control and operator submitted queries. The methodology of objective-driven monitoring for selective activation of sensors is implemented as a set of rules in the knowledge base of the monitor. Index Terms: Network, Quality of Service, Performance Management, Sensor, Monitoring, Knowledge-Based Systems. 1. Introduction Broadband networks consist of many subsystems (switching nodes, multiplexers, links, etc.) that are geographically distributed, carry multiple classes of traffic and have ...

This thesis addresses the problem of trying to locate the source of performance bottlenecks in large-scale parallel and distributed applications. Performance monitoring creates a dilemma: identifying a bottleneck necessitates collecting detailed information, yet collecting all this data can introduce serious data collection bottlenecks. At the same time, users are being inundated with volumes of complex graphs and tables that require a performance expert to interpret. I have developed a new approach that addresses both these problems by combining dynamic on-the-fly selection of what performance data to collect with decision support to assist users with the selection and presentation of performance data. The approach is called the W 3 Search Model. To make it possible to implement the W 3 Search Model, I have developed a new monitoring technique for parallel programs called Dynamic Instrumentation. The premise of my work is that not only is it possible to do on-line performance debu...

this paper is to determine the degree of invasiveness of PARMACS and how this invasiveness depends on the application, instrumentation and hardware platform. These results are compared to similar tools and a state of the art survey in this field is also presented.

Abstract. This paper presents HERCULE, an approach to non-invasively tracking end-user application activity in a distributed, component-based system. Such tracking can support the visualisation of user and application activity, system auditing, monitoring of system performance and the provision of feedback. A framework is provided that allows the insertion of proxies, dynamically and transparently, into a componentbased system. Proxies are inserted in between the user and the graphical user-interface and between the client application and the rest of the distributed, component-based system. The paper describes: how the code for the proxies is generated by mining component documentation; how they are inserted without affecting pre-existing code; and how information produced by the proxies can be used to model application activity. The viability of this approach is demonstrated by means of a prototype implementation. 1

I hereby declare that I am the sole author of this thesis. I authorize the University of Waterloo to lend thesis to other institutions or individuals for the purpose of scholarly research. B.R. Pekilis I further authorize the University of Waterloo to reproduce this thesis by photocopying or by any other means, in total or in part, at the request of others institution or individuals for the pur-pose of scholarly research. B.R. Pekilis ii The University of Waterloo requires the signatures of all persons using or photocopying this thesis. Please sign below, and give address and date. iii The focus of the research presented in this thesis is on session-oriented, soft real-time systems whose behavior is specified using the ITU Specification and Description Language

. This paper deals with performance evaluation of parallel and distributed systems based on monitoring concurrent interdependent activities. First, a model is introduced for describing the dynamic behavior of computer systems in terms of events. Then, the distributed hardware/hybrid monitor system ZM4 for event driven monitoring is presented. This monitor system can be adapted to arbitrary computer systems and can record event traces of arbitrary length. For monitoring distributed or large parallel systems it provides a global time base with a resolution of 100 ns. A tool environment is a prerequisite for successful performance evaluation. Therefore, the tool environment SIMPLE was developed. All tools use our data access interface TDL/POET which can decode measured data of arbitrary structure, format and representation. This makes evaluation independent of the monitor device(s) used and the system monitored. The data access interface also allows to merge data recorded by other monitor...